An essential role for retinoid signaling inanteroposterior neural specification and neuronaldifferentiationBruce Blumberg

Vitamin A and its derivatives, the retinoids, have beenimplicated as important signaling molecules in vertebratedevelopment, primarily due to numerous gain-of-functionexperiments involving treatment of embryos with exogenousretinoic acid. Collectively, the results from these experimentssuggested that retinoids were sufficient to mimic the action ofendogenous factors which posteriorize the embryonic axis,however the question of their necessity has remained unclear.Recent loss-of-function experiments from several laboratorieshave shown that retinoids are indispensable for regionalizingthe anteroposterior axis and promoting neuronal differ-entiation. These results firmly establish retinoids as endoge-nous signaling molecules that are required for embyronicpatterning.

Key words: loss-of-function / neuronal differentiation /posteriorization / RAR / retinoid

©1997 Academic Press Ltd

VITAMIN A IS REQUIRED for growth, vision, reproduc-tion, morphogenesis, hematopoiesis, immune func-tion, and differentiation of normal and malignanttissues.1 Although complete absence of vitamin A islethal, deficiency during development leads to aspectrum of well-characterized defects collectivelycalled the fetal vitamin A deficiency (VAD) syndrome.Depletion and repletion studies showed that all transRA (atRA) can substitute for most vitamin A require-ments leading to the hypothesis that atRA is the activeprinciple of vitamin A. While atRA can frequentlysupplant vitamin A, vision requires retinaldehyde,immune function requires 14-hydroxy-retroretinol2

and reproduction requires an, as yet, unidentifiedderivative. These findings, together with the discoveryof novel, bioactive retinoids in the embryo (describedbelow) suggest that vitamin A regulates a large

number of important biological processes via metabo-lism into a variety of known and unknown derivatives.Of particular interest is the requirement for retinoidsin specifying the vertebrate anteroposterior (A/P)axis, especially their potential role in patterning thecentral nervous system (CNS). Many excellent reviewsare available which treat various aspects of axialpatterning (especially in the mouse) and what roleretinoids might have in this process.3-8 This reviewdescribes recent evidence from loss-of-functionexperiments in chicken and Xenopus that establishesan indispensable role for retinoid signaling in pattern-ing the CNS and in promoting neuronaldifferentiation.

Retinoid signaling pathways

RA excess produces a spectrum of developmentaldefects affecting many of the same tissues as VAD, e.g.the heart, CNS, eyes, ears and reproductive tissues.Sensitivity of the same embryonic tissues to RAdeficiency or excess suggested that endogenous RAlevels required precise regulation for development toproceed correctly. A large body of evidence existslinking retinoids to the regulation of developmentallysignificant genes. Among these are genes involved indetermining positional identity along the anteropos-terior axis.

The biological effects of RA are thought to bemodulated primarily by the action of two classes ofligand activated transcription factors which are mem-bers of the nuclear hormone receptor superfamily(reviewed in refs 9, 10). The first types to bediscovered were the retinoic acid receptors (RARs)which bind to and are activated by all-trans retinoicacid (atRA) at nanomolar concentration.11,12 RARscomprise a family of three genes, the products ofwhich are three receptor subtypes. Each subtype hastwo or more isoforms which are the products ofdifferential promoter usage and alternative mRNAsplicing. Subsequently, a second class of RA receptor,

From the Gene Expression Laboratory, The Salk Institute forBiological Studies, 10010 North Torrey Pines Road, La Jolla, CA92037, USA

seminars in CELL & DEVELOPMENTAL BIOLOGY, Vol 8, 1997: pp 417–428

©1997 Academic Press Ltd1084-9521/97/040417 + 12 $25.00/0/sr970165

417

the retinoid ‘X’ receptors (RXRs) was identified.RXRs were shown to be activated by, but not to bindatRA.13 Retinoid ‘X’ was subsequently identified as9-cis RA (9cRA)14,15 and the RXR family was alsoshown to contain at least three genes.16 Interestingly,while atRA is specific for RARs, 9cRA is a pan agonistwhich binds to and activates RARs and RXRs with highaffinity. RAR requires heterodimerization with RXR inorder to bind its specific DNA response elementscalled RAREs (retinoic acid response elements) (Fig-ure 1A). In contrast, RXRs bind to RXREs as

homodimers (Figure 1B). Both types of receptorsregulate the transcription of response element-con-taining genes and provide a direct mechanism to linkretinoid concentration and gene expression.

All of the RAR and RXR genes have now beendisrupted in transgenic mice (ref 17, reviewed in ref18). The most dramatic phenotypes are seen in micewhere multiple receptor subtypes have been elimi-nated (see Ward and Morriss-Kay, this issue). Thephenotypes observed vary depending on the exactcombination of receptors missing. The knockouts

Figure 1. Multiple retinoid mediated signaling pathways (A) RAR and RXR heerodimerize to binddirect repeats separated by two or five nucleotides. In these heterodimers, RXR may bind ligand,depending on whether RAR is occupied and on the particular DNA-response element to which itis bound.23,87 (B)RXR can homodimerize on direct repeats separated by one nucleotide. Thesehomodimers are transcriptionally activated by binding 9cRA. (C) RXR is an obligate heterodimericpartner for many other ligand dependent receptors including PPAR, VDR, and TR. In addition,RXR partners with numerous orphan nuclear receptors (OR). In some permissive heterodimers(e.g. RXR:LXR) RXR is able to bind ligand and transduce 9cRA signals while in others (e.g.RXR:TR) it is transcriptionally inactive. Permissive heterodimers thus provide a third mechanismfor retinoid signaling to occur. (D) RXR can also partner with the orphan nuclear NGFI-B, whichordinarily binds to an extended half site sequence. RXR:NGFI-B heterodimers typically bind todirect repeats separated by five nucleotides and are transcriptionally activated by 9cRA.

B. Blumberg

418

recapitulate many aspects of fetal VAD but alsointroduce novel phenotypes such as homeotic verte-bral transformations and atavistic skeletal structures.18

There is an interesting general correlation betweenthe posteriorizing effects of RA excess and anteriorHox gene overexpression and between the anterioriz-ing effects of RAR compound knockouts and Hoxgene knockouts (reviewed in ref 5). These have alsopointed to an important role for the retinoid regula-tion of Hox gene expression in axial patterning.

In addition to the major retinoid response pathwaysit should be noted that RXR is a heterodimericpartner for a larger variety of known and orphannuclear receptors (reviewed in ref 19) (Figure 1C).This opens the possibility that additional retiniodeffects could be mediated via activation of RXR inpermissive heterodimeric complexes either alone, asexemplified by the activation of LXR:RXR hetero-dimers by 9cRA,20 or in combination with the ligandof the partner, as with RAR:RXR heterodimers.21 23

Synthetic compounds exist which selectively activateRXR in heterodimeric complexes24,25 hence it ispossible that endogenous compounds with similarselectivity also exist. Lastly RXR can modulate theactivity of at least two orphan nuclear receptors whichordinarily bind to DNA as monomers (NGFI-B,Nurr1) (Figure 1D).26,27 This multitude of retinoid-dependent signaling pathways underlies the ability ofretinoids to regulate diverse biological processes. Inaddition, all four types of pathways (Figure 1)function during embryogenesis which multiples thenumber of potential downstream genes under reti-noid control. This makes it important to identify theendogenous retinoids functioning duringdevelopment.

Endogenous bioactive retinoids

Endogenous bioactive retinoids have been describedin the early embryos of typical vertebrate modelsystems. atRA has been reported in the early embryosof zebrafish,28 chicken29 and mouse,30 although asdescribed below, some of the methods employed maynot be capable of discerning the identity of the activecompound. Endogenous, bioactive retinoids havebeen reported from Xenopus embryos, however, thereis some disagreement about which retinoids arepresent and at what amounts. To enable directcomparisons between differently reported results, Ihave used 1 µ1 and 1 mg as estimates of embryonicvolume and mass and recalculated the reported

concentrations. Durson and colleagues first reportedthe presence of ~ 150 nM atRA in the Xenopusneurula31 and later reported 4-oxo-tRA at ~ 20 nM.32

Chen et al33 used a reporter cell assay to identify totalembryonic retinoids and found ~ 110 nM of retinoicacid equivalent activity at the neurula stage. Sincetheir assay measured the biological activity of wholeembryo extracts, which contain numerous retiniods inaddition to retinoic acids, their finding should beconsidered an upper limit for the total RA-equivalentconcentration. Creech Kraft and colleagues reportedatRA and 9cRA in the early neurula each at approx-imately 120 ng/g which translates to nearly 800 nMtotal retinoic acid.34 We could not detect atRA or9cRA in the embryo but did identify oxo-retinoids atquantities ranging from 0–13 nM (4-oxo-retinol),50–144 nM (4-oxo-retinaldehyde), and 12–22 nM(4-oxo-RA) depending on the stage of developmentassayed.35 The total concentration of these three RARactivators in the neurula is ~ 92 nM, hence theyrepresent the bulk of the bioactive retinoids and thetotal amount is relatively close to that measured forwhole embryo extracts.33

The relatively large values obtained for retinoidswhich may not even be present in the embryoprobably derives from the method of identification.Compounds may be unambiguously identified byemploying multiple independent methods to charac-terize them. HPLC with single wavelength ultravioletabsorption detection was used in the two studieswhich gave high concentrations for atRA and9cRA.31,34 While this method may be sufficientlyaccurate for resolving mixtures of standards or identi-fying metabolites of added labeled compounds, thecomplex biochemical environment of the earlyembryo can easily lead to misidentification ofco-migrating compounds. In contrast, Pijnappel et al32

used HPLC with photodiode array detection (whichgives a complete UV absorbance profile rather than asingle wavelength readout) and two column systems toidentify 4-oxo-RA and obtained virtually identicalresults to Blumberg et al, who used HPLC, photodiodearray detection, multiple column systems, mass spec-troscopy and a sensitive bioassay to identifyretinoids.35

All three oxo-retinoids present in the Xenopusembryo are potent and selective activators forRARs.35,36 This led to the conclusion that 4-oxoderivatization, rather than representing a degradationpathway as had been previously assumed, results inthe conversion of retinol and retinal to bioactivehormones.35,36 This suggests that the spectrum of

Retinoids are required for neural patterning

419

bioactive retinoids is broader than previously sus-pected. The main conclusion from the study ofXenopus embryonic retinoids is that receptor agonistsare present in the embryo at the time when patterningis underway and at concentrations which stronglyactivate the receptors. This is also the time of maximalembryonic sensitivity to exogenous RA. Takentogether, these data support a potential role forretinoid signaling in early development.

Neural induction

The vertebrate nervous system is induced by signalsemanating from the dorsal mesoderm, or organizer,that divert the ectoderm away from an epidermal andtowards a neural fate. At the same time, signals fromthe organizer pattern the neural ectoderm along theA/P and D/V axis. The currently favored paradigmfor neural induction is the ‘activation-transformation’model (Figure 2).37,38 In this two-step model, ecto-derm is induced to form anterior neural tissue byvertical signals from the underlying mesoderm (acti-vation). Later, this anterior neural tissue is trans-formed to produce the full spectrum of anteropos-terior structures by both vertical signals from themesoderm and planar signals traveling through theectoderm (reviewed in refs 39, 40). The activation-transformation model is supported by experimentswhich showed that A/P pattern was progressivelyspecified after an initial induction to cement gland,which is considered the most anterior neural tissue inXenopus.41 Moreover, all direct neural inducersdescribed to date, namely, noggin, chordin, follistatin,the dominant negative activin type II and BMP-4 typeI receptors, induce neural tissue of anterior charac-ter.42-46 These findings all support a model where the

basal state of the neural ectoderm is anterior and thatadditional factors are required to generate the moreposterior parts of the nervous system. This two-stepprocess has been interpreted as a ‘two signal’ modelof neural induction.47 Much progress has been madein elucidating the identity and function of the‘activating’ factor, however, the identity of the endoge-nous transforming factor(s) has remained elusive.

What are the expected properties of a neuraltransforming factor? In order to be considered anendogenous transforming factor, the candidateshould possess several properties.48 First, it shouldhave the required activity, i.e. it should posteriorize.Second, it should be expressed when and wheretransformation occurs. Third, inhibition of its func-tion should anteriorize the embryo. Fourth, it shouldbe transmissible through tissue, both vertically and inthe plane of the ectoderm. Lastly, the degree oftransformation should ideally be concentration-dependent as has been suggested for the endogenoustransforming signal.40

RA was first proposed to be involved in neuraldevelopment when it was demonstrated that exoge-nously applied RA produces a concentration-depend-ent truncation of anterior, and enhancement ofposterior structures in Xenopus embryos31,41 throughits influence on the embryonic mesoderm and ecto-derm.49,50 At low concentrations, RA suppresses ante-rior and enhances posterior development within thehindbrain, without grossly affecting more anteriorparts of the brain51 while at high concentrations itcauses anterior truncations31,51 reflected by suppres-sion of anterior markers such as Otx2,52,53 Emxl,253

and Dlx1.53 In addition, atRA leads to an anterior shiftin the expression of many Hox genes (e.g. B1, B2, B3,B7).5,53-55 RA can also mimic the action of endoge-nous signals involved in inducing posterior geneexpression in the Xenopus nervous system.56

Figure 2. The activation transformation model of neural induction. In this model the mesodermproduces primary inducers which cause the overlying ectoderm to become anterior neural tissue.Then one or more transforming factors act to convert this anterior neural tissue into the completespectrum of positional values.

B. Blumberg

420

The hypothesis that RA is the endogenous agent ofposterior neural transformation31,56 was challenged byobservations which demonstrated a lack of coordinateposterior transformations by exogenous RA,51,57

inhibition of tail structures by RA,49 inconsistentresponse of posterior genes to exogenous RA applica-tion (e.g. no increase in the posterior marker Xhox3by RA)41,49,56,58 and no obvious positional shiftsresulting from localized application of RA.59 Themain problem in establishing what role retinoidsignaling truly plays in development is that all of theseexperiments were based on gain-of-function by addi-tion of exogenous RA. Hence while retinoids areclearly sufficient to produce effects similar to poster-iorization, there were, until recently, no experimentswhich address the question of its necessity in the sameprocesses.

Is retinoid signaling necessary and sufficientfor neural posteriorization?

In order to fully understand the role of retinoidsignaling in development it is necessary to performloss-of-function experiments. Several different strate-gies for achieving loss-of-function are possible. Theearliest and simplest test of retinoid function comesfrom depletion/repletion experiments. The embryois rendered vitamin A-deficient by dietary or othermeans, and either studied for developmental effectsor repleted with aRA at various times during develop-ment to assay the requirements for RA in particularprocesses. Despite the large body of literature con-cerning the fetal vitamin A deficiency (VAD) syn-drome, there has been relatively little publishedconcerning the effects of VAD on axial patterning. Anexciting recent result was published by Maden andcolleagues60 who showed that quail deficient invitamin A had considerable defects in axial pattern-ing. They detected alterations in the expressionpatterns of Hox-A2, Hox-B1, Hox-B4, Krox-20 and FGF-3which, taken together, led them to conclude that theposterior rhombomeres4-8 had been lost. In contrast,they detected no alterations in the expression ofdorsoventral markers, sonic hedgehog, islet-1 and Pax-3.This suggests that retinoid signaling is only requiredfor A/P and not D/V patterning.

The other strategies to attain loss-of-function relyon modulating the activity of retinoid receptors. Acurrent concept of how RARs function is illustrated inFigure 3A-C.61 A central feature is that RARs hetero-dimerize with RXRs and bind DNA in the absence of

ligand. In the unliganded state these heterodimersrecruit co-repressor proteins and actively suppresstranscription (Figure 3A). Upon the binding of anagonistic ligand, the receptor undergoes a conforma-tional change which leads to co-repressor release andbasal transcription levels (Figure 3B). The ligandedreceptor is now free to recruit co-activator proteinswhich, together with the transcriptional machinery,leads to activated transcription (Figure 3C). It ispossible to interfere with these processes at varioussteps and it could be presumed that different modesof interference would result in qualitatively differentresults.

Retinoid antagonists interfere with the binding ofRA or other agonistic ligands and are assumed toprevent the formation of the activated transcriptioncomplex (Figure 3D). In addition, one could spec-ulate that compounds exist which stabilize therepressed complex rather than just preventing forma-tion of the activated complex, virtually locking thesystem in the repressed state. This type of antagonistactually exists and is termed an ‘inverse agonist’ todistinguish it from the ‘neural antagonists’ whichdisrupt agonist binding but do not repress basaltranscription.62 Similarly, the ability of agonists todestabilize the repressed complex and stabilize theactivated complex varies and corresponds, in part, todifferences in potency of activation. The agonist/antagonist approach is quite specific for the pathwayunder study, however, localized effects are difficult orimpossible to achieve. RAR antagonists have beenused to great advantage in resurrecting a role for RAin limb patterning63 but have not yet been widelyemployed in A/P patterning studies. Carrasco andcolleagues have tested the ability of an RAR antagonistto alter the expression of Hox genes in chicken andXenopus (ref 55, A. Carrasco, personal communica-tion). They showed that while RA treatment led to ananterior expansion of Hox-B7 and C6 expression inchicken and Xenopus (posteriorization), treatmentwith the antagonist reduced the levels and extent ofHox-B7 and C6 expression (anteriorization).55 Theexpression of the hindbrain marker Krox-20 wasshifted anteriorly by RA and posteriorly by theantagonist which is also consistent with RA function-ing to posteriorize the axis (A. Carrasco, personalcommunication).

Because RARs require dimerization with RXR inorder to function, it is possible to modulate theiractivity by interfering with dimerization as originallyproposed by Herskowitz.64 One way to create such a‘dominant negative’ receptor is by overexpressing a

Retinoids are required for neural patterning

421

truncated RAR or another receptor which can seques-ter RXR in complexes which are incapable of bindingto RAREs (Figure 3E). Since the resulting dimer ismerely nonfunctional, it should be possible to recon-stitute normal activity by adding back an excess of

functional partner or wild type receptor. A potentialproblem with this approach is that RXR is a requireddimeric partner for numerous other receptors pre-sent during development and so one could havedifficulty distinguishing between effects resulting

Figure 3. Models for RAR function. (A–C) Three activation states for RXR:RAR heterodimers. (A)In the absence of ligand, RARs bind to DNA and recruit transcriptional corepressors that activelysuppress transcription from the response element. (B) Ligand binding induces a conformationalchange which triggers co-repressor dissociation and allows basal transcription levels. (C) Theliganded heterodimer is now free to recruit transcriptional coactivators such as SRC-1, CBP/P-300and other factors (CoAct-X). This results in the formation of an activated transcription complexand maximal expression from the response element. (D–F) Strategies for interfering with retinoidreceptor function. (D) RAR antagonists prevent the binding of ligands to the RAR, disruptformation of the activated transcription complex and result in transcriptional repression from theresponse element. Repression can be relieved by treatment with excess agonist. (E) Otherreceptors which interfere with the ability of RXR to heterodimerize with RAR can act in adominant negative fashion. Among these are RAR-ligand binding domain alone or other RXRpartners such as TR, v-erbA or COUP-TF. Since the target is RXR, this type of dominant negativerepression can be rescued by overexpressed RXR. (F) Another type of dominant negative receptoris caused by removing the activation domain from the carboxyl terminus of RARs. This prevents theconformational change normally elicited by ligand binding and results in dominant, constitutiverepression from the response element. Due to the similarity between the repression elicited byunliganded wild type receptors and the activation domain truncations, it is unlikely that this typeof dominant negative can be rescued by wild type receptor (which also represses) or ligand (whichis unable to dissociate corepressor).

B. Blumberg

422

from inhibition of RAR and those which derive frominhibiting another RXR-dependent pathway.

Several studies employed this type of dominantnegative approach. Overexpressed RARγ ligand bind-ing domain was microinjected into embryos andshown to block, in part, the teratogenic effects ofadded RA.65 Others used the thyroid hormone recep-tor or its mutant derivative v-erbA to sequester RXRand demonstrated reduced susceptibility toRA-teratogenesis and decreased induction ofRA-response genes such as Xlim-1 and Hox-D1.66-68

v-erbA injection was also able to reduce the expressionof Hox-D1 in whole embryos.68 Microinjection ofanother nuclear receptor, COUP-TFI, which likelyfunctions similarly, was shown to impair the inductionof RA-responsive genes and also reduced expressionof the hindbrain marker Krox-20.69

An alternative way to create dominant negativereceptors relies on the interaction between thereceptor and co-repressor protein(s) (Figure 3F).Several groups have shown that receptors lacking thecarboxyl-terminal activation domain behave as domi-nant transcriptional repressors. Although these recep-tors are capable of binding ligand70,71 the lack of theactivation domain prevents the conformationalchange which leads to co-repressor dissociation.72

Thus this type of dominant negative receptor simulta-neously sequesters RXR and ligand while constitu-tively interacting with co-repressor proteins to blocktranscription from RA target genes, even in thepresence of RA or other receptor agonists. Since thewild type receptor also interacts with co-repressorproteins it would not be expected to rescue this typeof dominant negative receptor.

We employed this type of dominant negativereceptor, together with a constitutively active receptorcreated by fusing the wild type receptor with thestrong transactivation domain from the herpes sim-plex virus protein VP-1673 to evaluate the result ofdominant gain or loss of retinoid signaling.74 This isequivalent to a local increase or decrease in retinoidlevels. These constructs were microinjected unilat-erally into Xenopus embryos and the effects on A/Pmarker genes assayed. As would be predicted ifretinoid signaling is required for posteriorization, thedominant negative receptor led to expansion of theforebrain and anterior midbrain as measured byposterior expansion of the forebrain marker Otx-2,and a posterior coordinate shift in the expressiondomains of en-2 and Krox-20 (Figure 4). Also con-sistent with an endogenous posteriorizing role forretinoid signaling, the expression of posterior mark-

ers Xlim-1, N-tubulin and Hox-B9 was reduced oreliminated (Figure 4). Interestingly, while the rhom-bomere 3 band of Krox-20 was shifted toward theposterior, the more posterior rhombomere 5 band waseliminated (Figure 4). Increased retinoid signaling,mediated by expression of the constitutively activereceptor, led to the opposite effect. These poster-iorized embryos were characterized by decreasedforebrain size (and a reduced Otx-2 domain) and ananterior coordinate shift, and a decrease in theexpression levels of en-2 and Krox-20 (Figure 4).Others used a different dominant negative RARisoform and showed loss of the posterior marker Hox-D1, and compression of Krox-20 expression in thehindbrain although no changes in Otx or Hox-B9were noted (H. Sive, personal communication). The

Figure 4. Anteroposterior positional shifts caused by local-ized perturbation of retinoid signaling. The leftmost set ofbars depicts the normal expression domains of some usefulA/P markers. Otx-2 marks the forebrain and anteriormidbrain. En-2 marks the border between the midbrain andhindbrain. Krox-20 marks rhombomeres 3 and 5 of thehindbrain while Hox-B9 is a spinal cord marker. Injection ofthe constitutively active receptor (which locally increasesretinoid signaling) leads to reduction or loss of Otx-2,rostral shifts and decreases in the expression domains ofEn-2 and Krox-20. In about 50% of the embryos, en andKrox-20 staining is lost altogether. There is no discernableeffect on Hox-B9 expression. In contrast, injection of thedominant negative receptor (which locally decreases reti-noid signaling) leads to caudal expansion of Otx-2 andcaudal shifts in the expression of en-2 and the r3 band ofKrox-20 expression. The r5 Krox-20 band is always lost andHox-B9 expression is greatly reduced or eliminated.

Retinoids are required for neural patterning

423

conclusion from these results is that retinoid signalingthrough xRARs is essential to correctly restrict theexpression of anterior genes and to enable theexpression of posterior marker genes. Hence reti-noids are excellent candidates to comprise oneimportant part of the posteriorizing signal. Thehindbrain appears to be extremely sensitive to thelevels of retinoid signaling as was also shown by thevitamin A deficiency experiments.

Retinoid signaling andactivation/transformation

What is the relationship between these new resultsand the status of the posteriorizing signal in theactivation transformation model? It had been pre-viously shown that basic fibroblast growth factor(bFGF) could posteriorize anterior neuroectoderm,in vitro and many inferred an endogenous role inneural induction and patterning (refs 75-77, reviewedin 47). Others showed that eFGF could posteriorizethe axis via induction of downstream genes Xcad3 andHox-A7 in vivo.78 Because eFGF is appropriatelyexpressed (notochord and posterior mesoderm) andinhibition of FGF signaling via overexpression of thedominant negative FGF receptor, XFD, reduced theexpression of the posterior markers Hox-A7 andXcad-3 but not the anterior markers Hox-B1 and Otx-2 it was suggested that eFGF is a good candidate for atleast part of the posterior transforming signal.78 Incontrast to these results, others demonstrated thattransgenic embryos expressing XFD (and non-mosa-ically deficient in FGF signaling) showed stronginhibition of posterior mesoderm but only limitedeffects on A/P patterning of the nervous system,suggesting that signaling through FGF receptors is notessential for neural posteriorization.79 In support ofthis contention it was recently shown that RA couldposteriorize anterior neuroectoderm injected withXFD while FGF could not.80 Thus, both retinoid andFGF signaling can posteriorize anterior neural tissuein vitro.

As described above, we showed that blockingendogenous retinoid signaling alone is sufficient toablate three posterior neural markers in vivo(although the embryos do have tails).74 Maden et alshowed that retinoid deficient quail embryos hadsevere patterning defects in the posterior hindbrain.60

The simplest interpretation of these results is thatneither retinoid signaling nor FGF signaling alone issufficient to posteriorize both the embryonic meso-

derm and neuroectoderm, in vivo. It could well bethat FGF and retinoids act synergistically to promoteposteriorization as was suggested previously based ontransplantation experiments.58 It is also not excludedthat signaling through other RARs and RXRs or as yetunidentified FGF-like receptors could be involved inrestricting anterior, and promoting posterior develop-ment. Additional factors such as members of thevertebrate hedgehog or Wnt gene families could alsoplay a role in posteriorization.81 More experimentswill be required to determine the precise role of eachcomponent, however, it seems clears that retinoidscomprise an indispensable part of the posteriorizingsignal, in vivo.

Retinoid signaling and neuronal differentiation

Retinoids were hypothesized to play an important rolein the development of the CNS primarily due to theability of RA excess or retinoid deficiency to elicitserious patterning defects. Additional evidence comesfrom cell culture studies which showed that atRAtreatment could alter neural differentiation of embry-onal carcinoma cells (reviewed in ref 3). Treatmentwith atRA induces the differentiation of P19 ter-atocarcinoma cells to neurons and astroglia82,83 andcauses cultured neuroblastoma cells to produce struc-tures resembling dendrites whereas other differ-entiating treatments lead to the formation of struc-tures resembling axons.84 In addition, retinoidreceptors, cellular binding proteins and bioactiveretinoids are all found in the developing CNS.3

Two lines of evidence support an endogenous rolefor RA in neuronal differentiation — depletionstudies and dominant negative interference experi-ments. Maden et al showed that retinoid-deficientquail embryos are lacking in neural crest and fail toextend neurites into the periphery.60 More specifi-cally, it appears that neural crest cells are specified butfail to survive and migrate. In addition, those fewneurons which did remain often followed chaoticmigration paths, suggesting that positional identitieswere incorrect. More recent data from the samelaboratory suggests that retinoids influence apoptosisof neuronal precursors in the chick. Vitamin Adeficient quail embryos show a band of apoptosis firstin the underlying mesoderm and later in the regionfated to form rhombomeres 4–8 of the hindbrain (M.Maden, personal communication). Thus it appearsthat incorrectly patterned cells fail to survive.

B. Blumberg

424

Two studies using dominant negative RARs showedthat signaling through xRARα is required for neu-ronal differentiation.74,85 N-tubulin is a molecularmarker for primary neurons, which normally differ-entiate in three stripes on either side of the dorsalmidline at the neural plate stage and later give rise tosensory, motor and interneurons.86 Microinjection ofan xRARα1 dominant negative construct was able toprevent the expression of N-tubulin in the Xenopusneurula. It was concluded that retinoid signaling isrequired for differentiation of all three major classesof neurons since all three stripes of N-tubulin areaffected by the dominant negative RAR.74 Others useda different dominant negative receptor isoform,xRARα2·2, to show that the formation of sensoryneurons is dependent on retinoid signaling as meas-ured by expression of the neuronal markers XIF-3 andHNK-1.85 Microinjection of the dominant negativereceptor led to a dose-dependent decrease in thenumber of sensory neurons. Furthermore, micro-injection of mRNA encoding the wild type xRARα2·2together with xRXRâ led to the formation of ectopicneurons adjacent to the neural tube.85 Additionalexperiments with inhibitors of cell proliferationexcluded the possibility that the ectopic neuronsformed as a result of increased proliferation ofneuronal precursors, however, the mechanism bywhich they are formed remains unclear. Furtherexperiments will be required to determine whetherthese ectopic neurons form beyond the boundaries ofthe neural plate or instead form within the neuralplate but are not incorporated into the neural tube.The overall conclusion is that retinoid signaling playsa quantitative role in determining the number ofprimary neurons.

Future directions

As described above, the inescapable conclusion is thatretinoid signaling is required for correct A/P neuralspecification and neuronal differentiation. Retinoidsand their receptors are localized in the correct timeand at the correct place to act as posteriorizingfactors. Most importantly, inhibition of retinoid sig-naling interferes with patterning in a predictable way— posterior marker genes are lost and anterior genesextend more caudally than usual. Inhibition ofretinoid signaling interferes with neuronal differ-entiation while excess signaling leads to the formationof ectopic neurons. The reciprocal nature of thephenotypes from increased or decreased signaling is

completely consistent with a role for retinoids in thepatterning process itself. However, many questionsremain to be answered.

How exactly are retinoids and their receptorsfunctioning to mediate patterning? What are thecritical target genes? Is there, in fact, a gradient ofbioactive retinoids from posterior to anterior and doretinoids act as gradient morphogens to specifyposition along the A/P axis? Do retinoids diffuse froma discrete source in the embryo or are there multiplesources at different times of development? How manyfactors are important for A/P patterning? Both FGFand retinoid signaling appear to have important rolesin posteriorization of the axis (Figure 5). FGF seemsmore important for mesodermal patterning whereasretinoids seem more important for neural patterning.It is possible, even likely, that these two pathwaysinteract in the patterning process, hence one wouldlike to know the mode and extent of such inter-actions. When do the retinoids act (Figure 5)? Bothbioactive retinoids and functional receptors are pre-sent in the Xenopus egg hence there is no limit on howearly the pathway could be activated.

Alterations in A/P neural patterning could resultfrom changes in the underlying mesoderm, reflect arequirement for retinoid signaling in the ectoderm,or both. Results from gain-of-function studies suggestthat both are important. Are retinoids acting on theinducing signals or the responding tissues? Localizedreceptor expression in defined regions of the embryo,combined with the types of embryological manipula-tions possible in Xenopus will go a long way towardclarifying these issues. Evaluating the responses of anextensive set of molecular markers in the manipulatedXenopus embryo and integrating this information withthe available data from mouse and chicken embryos,should provide important insights into the processesunderlying patterning. As Hox genes have alreadybeen shown to play important roles in A/P patterning,it will be critical to understand how they are regulatedand which genes are downstream of retinoid recep-tors and upstream of Hox.

What is the role of retinoid signaling in neuronaldifferentiation? How does interfering with signalinglead to underproduction of neurons while excesssignaling gives ectopic neurons? Added exogenousRA increases both the width of the neural plate andthe number of primary neurons whereas inhibition ofretinoid signaling drastically reduces the number ofprimary neurons but does not appear to alter neuralplate boundaries much. It is likely that retinoids aremodulating the activity of certain proneural genes

Retinoids are required for neural patterning

425

which, in turn influence the number of primaryneurons formed. The timing of neuronal differ-entiation is also linked to A/P patterning which mayplay a role in the ultimate number of neurons.52

Understanding the role of retinoid signaling in thetiming of A/P neural patterning and neuronal differ-entiation identifying possible connections betweenapoptosis and incorrect patterning will be of consider-able interest.

Retinoids enjoyed a period of great prominence inembryological thinking during the 1980s only tobecome controversial, or even dismissed, as importantpatterning molecules in recent years. With the resultsof the past year or two it is now firmly established thatretinoids are necessary for correct anterior neuralpatterning, necessary and sufficient for posteriorneural patterning (especially in the hindbrain) andindispensable for neuronal differentiation in vivo.While many important questions remain, it seemsclear that the search for answers will greatly illuminate

our understanding of how the vertebrate body axis isformed.

Acknowledgements

I thank Nancy Papalopulu and Juan Carlos Izpisua Bel-monte for comments on the manuscript and AndresCarrasco, Malcom Maden, Hazel Sive and You Katsuyamafor generously sharing results prior to publication. Iespecially wish to thank Ron Evans for his continuoussupport and encouragement of this work. Supported bygrants from the NIH (GM-26444) and the G. Harold andLeila Y. Mathers Foundation (to Ron Evans).

References

1. Sporn MB, Roberts AB, Goodman DS (1994) The Retinoids:Biology, Chemistry, and Medicine. 2nd Ed.. Raven Press, NewYork

2. Buck J, Derguini F, Levi E, Nakanishi K, Hammerling U (1991)Intracellular signaling by 14-hydroxy-4,14-retro-retinol. Science254:1654-1565

Figure 5. Models for posteriorization. (A) Retinoids, FGFs and perhaps other factors may act onindependent pathways to achieve posteriorization of the CNS. (B) Retinoid and FGF signalingpathways may interact or synergize in posteriorizing the CNS. (C) Posteriorization may be amultiple step process wherein the axis is first crudely subdivided into anterior and posteriordomains and then later refined into the full spectrum of A/P values.

B. Blumberg

426

3. Maden M, Holder N (1992) Retinoic acid and the developmentof the central nervous system. BioEssays 14:431-438

4. Yamada T (1994) Caudalization by the amphibian organizer:brachyury, convergent extension and retinoic acid. Develop-ment 120:3051-3062

5. Conlon RA (1995) Retinoic acid and pattern formation invertebrates. Trends Genet 11:314-319

6. Lumsden A, Krumlauf R (1996) Patterning the vertebrateneuraxis. Science 274:1109-1115

7. Joyner AL (1996) Engrailed, Wnt and Pax genes regulatemidbrain-hindbrain development. Trends Genet 12:15-20

8. Bally-Cuif L, Boncinelli E (1997) Transcription factors andhead formation in vertebrates. BioEssays 19:127-135

9. Mangelsdorf DJ, Umesono K, Evans RM (1994) The retinoidreceptors, in The Retinoids: Biology, Chemistry and Medicine(Sporn MB, Roberts AB, Goodman DS, eds) pp 319–349 2ndEd. Raven Press, New York, NY

10. Chambon P (1996) A decade of molecular biology of retinoicacid receptors. FASEB J 10:940-954

11. Petkovich M, Brand NJ, Krust A, Chambon P (1987) A humanretinoic acid receptor which belongs to the family of nuclearreceptors. Nature 330:444-450

12. Giguere V, Ong ES, Segui P, Evans RM (1987) Identification ofa receptor for the morphogen retinoic acid. Nature330:624-629

13. Mangelsdorf DJ, Ong ES, Dyck JA, Evans RM (1990) Nuclearreceptor that identifies a novel retinoic acid response pathway.Nature 345:224-229

14. Levin AA, Sturzenbecker LJ, Kazmer S, Bosakowski T, HuseltonC, Allenby G, Speck J, Kratzeisen C, Rosenberger J, Lovey A,Grippo JF (1992) 9-cis retinoic acid stereoisomer binds andactivates the nuclear receptor RXRα. Nature 355:359-361

15. Heyman RA, Mangelsdorf DJ, Dyck JA, Stein RB, Eichele G,Evans RM, Thaller C (1992) 9-cis retinoic acid is a high affinityligand for the retinoid X receptor. Cell 68:397-406

16. Mangelsdorf DJ, Borgmeyer U, Heyman RA, Zhou J, Ong ES,Oro A, Kakizuka A, Evans RM (1992) Characterization of threeRXR genes that mediate the action of 9-cis retinoic acid. GenesDev 6:329-344

17. Krezel W, Dupe V, Mark M, Dierich A, Kastner P, Chambon P(1996) RXRg null mice are apparently normal and compoundRXRα+/–/RXRâ–/–/RXRγ–/– mutant mice are viable. Proc NatlAcad Sci (USA) 93:9010-9014

18. Kastner P, Mark M, Chambon P (1995) Nonsteroid nuclearreceptors: what are genetic studies telling us about their role inreal life. Cell 83:859-869

19. Mangelsdorf DJ, Evans RM (1995) The RXR heterodimers andorphan receptors. Cell 83:841-850

20. Willy PJ, Umesono K, Ong ES, Evans RM, Heyman RA,Mangelsdorf DJ (1995) LXR, a nuclear receptor that defines adistinct retinoid-response pathway. Genes Dev 9:1033-1045

21. Minucci S, Leid M, Toyama R, Saint-Jeannet J-P, Peterson VJ,Horn V, Ishmael JE, Bhattacharyya N, Dey A, Dawid IB, OzatoK (1997) Retinoid X receptor (RXR) within the RXR-retinoicacid receptor heterodimer binds its ligand and enhancesretinoid-dependent gene expression. Mol Cell Biol 17:644-655

22. Roy B, Taneja R, Chambon P (1995) Synergistic activation ofretinoic acid (RA)-responsive genes and induction of embry-onal carcinoma cell differentiation by an RA receptor alpha(RARα)-, RARâ-, or RARγ-selective ligand in combination witha retinoid X receptor-specific ligand. Mol Cell Biol15:6481-6487

23. Forman BM, Umesono K, Chen J, Evans RM (1995a) Uniqueresponse pathways are established by allosteric interactionsamong nuclear hormone receptors. Cell 81:541-550

24. Lala DS, Mukherjee R, Schulman IG, Koch SS, Dardashti LJ,Nadzan AM, Croston CE, Evans RM, Heyman RA (1996)Activation of specific RXR heterodimers by an antagonist ofRXR homodimers. Nature 383:450-453

25. Schulman IG, Li C, Schwabe JWR, Evans RM (1997) Thephantom ligand effect: allosteric control of transcription by theretinoid X receptor. Genes Dev 11:299-308

26. Forman BM, Goode E, Chen J, Oro AE, Bradley DJ, PerlmannT, Noonan DJ, Burka LT, McMorris T, Lamph WW, Evans RM,Weinberger C (1995) Identification of a nuclear receptor that isidentified by farnesol metabolites. Cell 81:687-693

27. Perlmann T, Jansson L (1995) A novel pathway for vitamin Asignaling mediated by RXR heterodimerization with NGFI-Band NURR1. Genes Dev 9:769-782

28. Costaridis P, Horton C, Zeitlinger J, Holder N, Maden M (1996)Endogenous retinoids in the zebrafish embryo and adult. DevDynam 205:41-51

29. Chen Y, Huang L, Russo AF, Solursh M (1992) Retinoic acid isenriched in Hensen’s node and is developmentally regulated inthe early chicken embryo. Proc Natl Acad Sci (USA)89:10056-10059

30. Horton C, Maden M (1995) Endogenous distribution ofretinoids during normal development and teratogenesis in themouse embryo. Dev Dynam 202:312-323

31. Durston AJ, Timmermans JP, Hage WJ, Hendriks HF, de VriesNJ, Heideveld M, Nieuwkoop PD (1989) Retinoic acid causesan anteroposterior transformation in the developing centralnervous system. Nature 340:140-144

32. Pijnappel WW, Hendriks HF, Folkers GE, van den Brink CE,Dekker EJ, Edelenbosch C, van der Saag PT, Durston AJ (1993)The retinoic ligand 4-oxo-retinoic acid is a highly activemodulator of positional specification. Nature 366:340-344

33. Chen YP, Huang L, Solursh M (1994) A concentration gradientof retinoids in the early Xenopus laevis embryo. Dev Biol161:70-76

34. Creech Kraft J, Schuh T, Juchau M, Kimelman D (1994) Theretinoid X receptor ligand, 9-cis retinoic acid, is a potentialregulator of early Xenopus development. Proc Natl Acad Sci(USA) 91:3067-3071

35. Blumberg B, Bolado J Jr, Derguini F, Craig AG, Moreno TA,Charkravarti D, Heyman RA, Buck J, Evans RM (1996) NovelRAR ligands in Xenopus embryos. Proc Natl Acad Sci (USA)93:4873-4878

36. Achkar C, Derguini F, Blumberg B, Langston A, Levin AA,Speck J, Evans RM, Bolado J Jr, Buck J, Gudas LJ (1996) 4-oxo-retinol, a new natural ligand and transactivator of the retinoicacid receptors. Proc Natl Acad Sci (USA) 93:4879-4884

37. Nieuwkoop PD (1952) Activation and organization of thecentral nervous system in amphibians. Part III. synthesis of anew working hypothesis. J Exp Zool 120:83-108

38. Eyal-Giladi H (1954) Dynamic aspects of neural induction.Arch Biol 65:180-259

39. Saxen L (1989) Neural induction. Int J Dev Biol 33:21-4840. Gilbert SF, Saxen L (1993) Spemann’s organizer: models and

molecules. Mech Dev 41:73-8941. Sive HL, Draper BW, Harland R, Weintraub H (1990) Identi-

fication of retinoic acid-sensitive period during primary axisformation in Xenopus laevis. Genes Dev 4:932-942

42. Lamb TM, Knecht AK, Smith WC, Stachel SE, Economides AN,Stahl N, Yancopolouos GD, Harland RM (1993) Neuralinduction by the secreted polypeptide noggin. Science262:713-718

43. Sasai Y, Lu B, Steinbeisser H, De Robertis EM (1995)Regulation of neural induction by the Chd and Bmp-4antagonistic patterning signals in Xenopus. Nature 376:333-336

44. Hemmati-Brivanlou A, Kelly OG, Melton DA (1994) Follistatin,an antagonist of activin, is expressed in the Spemann organizerand displays direct neuralizing activity. Cell 77:283-295

45. Hemmati-Brivanlou A, Melton DA (1994) Inhibition of activinreceptor signaling promotes neuralization in Xenopus. Cell77:273-281

46. Hawley SH, Wunnenberg-Stapleton K, Hashimoto C, LaurentMN, Watabe T, Blumberg G, Cho KWY (1995) Disruption of

Retinoids are required for neural patterning

427

BMP signals in embryonic Xenopus ectoderm leads to directneural induction. Genes Dev 9:2923-2935

47. Doniach T (1995) Basic FGF as an inducer of anteroposteriorneural pattern. Cell 83:1067-1070

48. Slack JMW (1993) Embryonic Induction. Mech Dev 41:91-10749. Ruiz i Altaba A, Jessell T (1991) Retinoic acid modifies

mesodermal patterning in early Xenopus embryos. Genes Dev5:175-187

50. Sive HL, Cheng PF (1991) Retinoic acid perturbs the expres-sion of Xhox-lab genes and alters mesodermal determinationin Xenopus laevis. Genes Dev 5:1321-1332

51. Papalopulu N, Clark JDW, Bradley L, Wilkinson D, Krumlauf R,Holder N (1991) Retinoic acid causes abnormal developmentand segmental patterning of the anterior hindbrain in Xenopusembryos. Development 113:1145-1158

52. Papalopulu N, Kinter C (1996) A posteriorizing factor, retinoicacid, reveals that anteroposterior patterning controls thetiming of neuronal differentiation in Xenopus neurectoderm.Development 122:3409-3418

53. Simeone A, Avantaggiato V, Moroni MC, Mavilio F, Arra C,Cotelli F, Nigro V, Acampora D (1995) Retinoic acid inducesstage-specific antero-posterior transformation of rostral centralnervous system. Mech Dev 51:83-98

54. Conlon RA, Rossant J (1992) Exogenous retinoic acid rapidlyinduces anterior ectopic expression of murine Hox-2 genes invivo. Development 116:357-368

55. Lopez SL, Dono R, Zeller R, Carrasco AE (1995) Differentialeffects of retinoic acid and a retinoid antagonist on the spatialdistribution of the homeoprotein Hoxb-7 n vertebrateembryos. Dev Dynam 204:457-471

56. Sharpe CR (1991) Retinoic acid can mimic endogenous signalsinvolved in transformation of the Xenopus nervous system.Neuron 7:239-247

57. Holder N, Hill J (1991) Retinoic acid modifies development ofthe midbrain-hindbrain border and affects cranial ganglionformation in zebrafish embryos. Development 113:1159-1170

58. Cho KWY, De Robertis EM (1990) Differential activation ofXenopus homeobox genes by mesoderm inducing growthfactors and retinoic acid. Genes Dev 4:1910-1916

59. Drysdale TA, Crawford MJ (1994) Effects of localized applica-tion of retinoic acid on Xenopus laevis development. Dev Biol162:394-401

60. Maden M, Gale W, Kostetskii I, Zile M (1996) VitaminA-deficient quail embryos have half a hindbrain and otherneural defects. Curr Biol 6:417-426

61. Chakravarti D, LaMorte VJ, Nelson MC, Nakajina T, SchulmanIG, Juguilon H, Montminy M, Evans RM (1996) Role of CBP/P300 in nuclear receptor signaling. Nature 383:99-103

62. Klein ES, Pino NE, Johnson AT, Davies PJ, Nagpal S, ThacherSM, Krasinski G, Chandraratna RAS (1996) Identification andfunctional separation of retinoic acid receptor neutral antago-nists and inverse agonists. J Biol Chem 271:22691-22696

63. Helms JA, Kim CH, Eichele G, Thaller C (1996) Retinoic acidsignaling is required during early chick limb development.Development 122:1385-1394

64. Herskowitz I (1987) Functional inactivation of genes bydominant negative mutations. Nature 329:219-222

65. Smith DP, Mason CS, Jones E, Old R (1994) Expression of adominant negative retinoic acid receptor γ in Xenopus embryosleads to partial resistance to retinoic acid. Roux’s. Arch DevBiol 203:254-265

66. Banker DE, Eisenman RN (1993) Thyroid hormone receptorcan modulate retinoic acid-mediated axis formation in frogembryogenesis. Mol Cell Biol 13:740-752

67. Schuh TJ, Hall BL, Kraft JC, Privalsky ML, Kimelman D (1993)v-erbA and citral reduce the teratogenic effects of all-transretinoic acid and retinol, respectively, in Xenopus embryos.Development 119:785-798

68. Kolm PJ, Sive HL (1995) Regulation of the Xenopus labialhomeodomain genes, HoxA1 and HoxD1: activation by reti-noids and peptide growth factors. Dev Biol 167:34-49

69. Schuh TJ, Kimelman D (1995) COUP-TFI is a potentialregulator of retinoic acid-modulated development in Xenopusembryos. Mech Dev 51:39-49

70. Allenby G, Bocquel M, Saunders M, Kazmer S, Speck J,Rosenberger M, Lovey A, Kastner P, Grippo JF, Chambon P,Levin AA (1993) Retinoic acid receptors and retinoic Xreceptors: Interactions with endogenous retinoic acids. ProcNatl Acad Sci (USA) 90:30-34

71. Damm K, Evans RM (1993) Identification of a domain requiredfor oncogenic activity and transcriptional suppression by v-erbAand thyroid-hormone receptor alpha. Proc Natl Acad Sci (USA)90:10668-10672

72. Chen JD, Umesono K, Evans RM (1996) SMRT isoformsmediate repression and anti-repression of nuclear receptorheterodimers. Proc Natl Acad Sci (USA) 93:7567-7571

73. Sadowski I, Ma J, Triezenberg S, Ptashne M (1988) GAL4-VP16is an unusually potent transcriptional activator. Nature335:563-564

74. Blumberg B, Bolado J, Moreno TA, Kintner C, Evans RM,Papalopulu N (1997) An essential role for retinoid signaling inanteroposterior neural patterning. Development 124:373-379

75. Cox WG, Hemmati-Brivanlou A (1995) Caudalization of neuralfate by tissue recombination and bFGF. Development121:4349-4358

76. Kengaku M, Okamoto H (1995) bFGF as a possible morphogenfor the anteroposterior axis of the central nervous system inXenopus. Development 121:3121-3130

77. Lamb TM, Harland RM (1995) Fibroblast growth factor is adirect neural inducer, which combined with noggin generatesanterior-posterior neural pattern. Development 121:3627-3636

78. Pownall ME, Tucker AS, Slack JMW, Isaacs HV (1996) eFGF,Xcad3, and Hox genes form a molecular pathway that estab-lishes the anteroposterior axis in Xenopus. Development122:3881-3892

79. Kroll KL, Amaya E (1996) Transgenic Xenopus embryos fromsperm nuclear transplantations reveal FGF signaling require-ments during gastrulation. Development 122:3173-3183

80. Bang AG, Papalopulu N, Kintner C, Goulding MD (1997)Expression of Pax-3 is initiated in the early neural plate byposteriorizing signals produced by the organizer and byposterior non-axial mesoderm. Development 124:2075-2085

81. McGrew LL, Lai CJ, Moon RT (1995) Specification of theanteroposterior neural axis through synergistic interaction ofthe Wnt signaling cascade with noggin and follistatin. Dev Biol172:337-342

82. Jones-Villeneuve EMV, McBurney MW, Rogers KA, Kalnins VI(1982) Retinoic acid induces embryonal carcinoma cells todifferentiate into neurons and glial cells. J Cell Biol94:253-262

83. McBurney MW, Jones-Villeneuve EMV, Edwards MKS, Ander-son PJ (1982) Control of muscle and neuronal differentiationin a cultured embryonal carcinoma cell line. Nature299:165-167

84. Sidell N (1982) Retinoic acid-induced growth inhibition andmorphologic differentiation of human neuroblastoma cells invitro. J Natl Cancer Inst 68:589-593

85. Sharpe CR, Goldstone K (1997) Retinoid receptors promoteprimary neurogenesis in Xenopus. Development 124:515-523

86. Chitnis A, Henrique D, Lewis J, Ish-Horowicz D, Kintner C(1995) Primary neurogenesis in Xenopus embryos regulated bya homologue of the Drosophila neurogenic gene Delta. Nature375:761-766

87. La Vista-Picard N, Hobbs PD, Pfahl M, Dawson MI, Pfahl M(1996) The receptor-DNA complex determines the retinoidresponse: a mechanism for the diversification of the ligandsignal. Mol Cell Biol 16:4137-4146

B. Blumberg

428